Support Giga AM transducer V2 (#2136)

.github/scripts/test-offline-ctc.sh

@@ -47,9 +47,23 @@ for type in base small; do
   rm -rf sherpa-onnx-dolphin-$type-ctc-multi-lang-2025-04-02
 done
 
+log "------------------------------------------------------------"
+log "Run NeMo GigaAM Russian models v2"
+log "------------------------------------------------------------"
+curl -SL -O https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/sherpa-onnx-nemo-ctc-giga-am-v2-russian-2025-04-19.tar.bz2
+tar xvf sherpa-onnx-nemo-ctc-giga-am-v2-russian-2025-04-19.tar.bz2
+rm sherpa-onnx-nemo-ctc-giga-am-v2-russian-2025-04-19.tar.bz2
+
+$EXE \
+  --nemo-ctc-model=./sherpa-onnx-nemo-ctc-giga-am-v2-russian-2025-04-19/model.int8.onnx \
+  --tokens=./sherpa-onnx-nemo-ctc-giga-am-v2-russian-2025-04-19/tokens.txt \
+  --debug=1 \
+  ./sherpa-onnx-nemo-ctc-giga-am-v2-russian-2025-04-19/test_wavs/example.wav
+
+rm -rf sherpa-onnx-nemo-ctc-giga-am-v2-russian-2025-04-19
 
 log "------------------------------------------------------------"
-log "Run NeMo GigaAM Russian models"
+log "Run NeMo GigaAM Russian models v1"
 log "------------------------------------------------------------"
 curl -SL -O https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/sherpa-onnx-nemo-ctc-giga-am-russian-2024-10-24.tar.bz2
 tar xvf sherpa-onnx-nemo-ctc-giga-am-russian-2024-10-24.tar.bz2

.github/scripts/test-offline-transducer.sh

@@ -15,6 +15,24 @@ echo "PATH: $PATH"
 
 which $EXE
 
+log "------------------------------------------------------------"
+log "Run NeMo GigaAM Russian models v2"
+log "------------------------------------------------------------"
+curl -SL -O https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/sherpa-onnx-nemo-transducer-giga-am-v2-russian-2025-04-19.tar.bz2
+tar xvf sherpa-onnx-nemo-transducer-giga-am-v2-russian-2025-04-19.tar.bz2
+rm sherpa-onnx-nemo-transducer-giga-am-v2-russian-2025-04-19.tar.bz2
+
+$EXE \
+  --encoder=./sherpa-onnx-nemo-transducer-giga-am-v2-russian-2025-04-19/encoder.int8.onnx \
+  --decoder=./sherpa-onnx-nemo-transducer-giga-am-v2-russian-2025-04-19/decoder.onnx \
+  --joiner=./sherpa-onnx-nemo-transducer-giga-am-v2-russian-2025-04-19/joiner.onnx \
+  --tokens=./sherpa-onnx-nemo-transducer-giga-am-v2-russian-2025-04-19/tokens.txt \
+  --model-type=nemo_transducer \
+  ./sherpa-onnx-nemo-transducer-giga-am-v2-russian-2025-04-19/test_wavs/example.wav
+
+rm sherpa-onnx-nemo-transducer-giga-am-v2-russian-2025-04-19
+
+
 log "------------------------------------------------------------------------"
 log "Run zipformer transducer models (Russian)                              "
 log "------------------------------------------------------------------------"

.github/workflows/export-nemo-giga-am-to-onnx.yaml

@@ -43,7 +43,8 @@ jobs:
           mv -v scripts/nemo/GigaAM/tokens.txt $d/
           mv -v scripts/nemo/GigaAM/*.wav $d/test_wavs/
           mv -v scripts/nemo/GigaAM/run-ctc.sh $d/
-          mv -v scripts/nemo/GigaAM/*-ctc.py $d/
+          mv -v scripts/nemo/GigaAM/export-onnx-ctc.py $d/
+          cp -v scripts/nemo/GigaAM/test-onnx-ctc.py $d/
 
           ls -lh scripts/nemo/GigaAM/
 
@@ -71,7 +72,8 @@ jobs:
           mv -v scripts/nemo/GigaAM/tokens.txt $d/
           mv -v scripts/nemo/GigaAM/*.wav $d/test_wavs/
           mv -v scripts/nemo/GigaAM/run-rnnt.sh $d/
-          mv -v scripts/nemo/GigaAM/*-rnnt.py $d/
+          mv -v scripts/nemo/GigaAM/export-onnx-rnnt.py $d/
+          cp -v scripts/nemo/GigaAM/test-onnx-rnnt.py $d/
 
           ls -lh scripts/nemo/GigaAM/
 
@@ -91,11 +93,12 @@ jobs:
           mkdir $d/test_wavs
           rm scripts/nemo/GigaAM/v2_ctc.onnx
           mv -v scripts/nemo/GigaAM/*.int8.onnx $d/
-          cp -v scripts/nemo/GigaAM/LICENCE $d/
+          cp -v scripts/nemo/GigaAM/LICENSE $d/
           mv -v scripts/nemo/GigaAM/tokens.txt $d/
           mv -v scripts/nemo/GigaAM/*.wav $d/test_wavs/
-          mv -v scripts/nemo/GigaAM/run-ctc.sh $d/
+          mv -v scripts/nemo/GigaAM/run-ctc-v2.sh $d/
           mv -v scripts/nemo/GigaAM/*-ctc-v2.py $d/
+          cp -v scripts/nemo/GigaAM/test-onnx-ctc.py $d/
 
           ls -lh scripts/nemo/GigaAM/
 
@@ -103,8 +106,36 @@ jobs:
 
           tar cjvf ${d}.tar.bz2 $d
 
+      - name: Run Transducer v2
+        shell: bash
+        run: |
+          pushd scripts/nemo/GigaAM
+          ./run-rnnt-v2.sh
+          popd
+
+          d=sherpa-onnx-nemo-transducer-giga-am-v2-russian-2025-04-19
+          mkdir $d
+          mkdir $d/test_wavs
+
+          mv -v scripts/nemo/GigaAM/encoder.int8.onnx $d/
+          mv -v scripts/nemo/GigaAM/decoder.onnx $d/
+          mv -v scripts/nemo/GigaAM/joiner.onnx $d/
+
+          cp -v scripts/nemo/GigaAM/*.md $d/
+          cp -v scripts/nemo/GigaAM/LICENSE $d/
+          mv -v scripts/nemo/GigaAM/tokens.txt $d/
+          mv -v scripts/nemo/GigaAM/*.wav $d/test_wavs/
+          mv -v scripts/nemo/GigaAM/run-rnnt-v2.sh $d/
+          cp -v scripts/nemo/GigaAM/test-onnx-rnnt.py $d/
+
+          ls -lh scripts/nemo/GigaAM/
+
+          ls -lh $d
+
+          tar cjvf ${d}.tar.bz2 $d
 
       - name: Release
+        if: github.repository_owner == 'csukuangfj'
         uses: svenstaro/upload-release-action@v2
         with:
           file_glob: true
@@ -114,7 +145,16 @@ jobs:
           repo_token: ${{ secrets.UPLOAD_GH_SHERPA_ONNX_TOKEN }}
           tag: asr-models
 
-      - name: Publish to huggingface (Transducer)
+      - name: Release
+        if: github.repository_owner == 'k2-fsa'
+        uses: svenstaro/upload-release-action@v2
+        with:
+          file_glob: true
+          file: ./*.tar.bz2
+          overwrite: true
+          tag: asr-models
+
+      - name: Publish to huggingface (CTC)
         env:
           HF_TOKEN: ${{ secrets.HF_TOKEN }}
         uses: nick-fields/retry@v3
@@ -126,11 +166,66 @@ jobs:
             git config --global user.email "csukuangfj@gmail.com"
             git config --global user.name "Fangjun Kuang"
 
+            d=sherpa-onnx-nemo-ctc-giga-am-russian-2024-10-24/
+            export GIT_LFS_SKIP_SMUDGE=1
+            export GIT_CLONE_PROTECTION_ACTIVE=false
+            rm -rf huggingface
+            git clone https://csukuangfj:$HF_TOKEN@huggingface.co/csukuangfj/$d huggingface
+            cp -av $d/* ./huggingface
+            cd huggingface
+            git lfs track "*.onnx"
+            git lfs track "*.wav"
+            git status
+            git add .
+            git status
+            git commit -m "add models"
+            git push https://csukuangfj:$HF_TOKEN@huggingface.co/csukuangfj/$d main
+
+      - name: Publish to huggingface (Transducer)
+        env:
+          HF_TOKEN: ${{ secrets.HF_TOKEN }}
+        uses: nick-fields/retry@v3
+        with:
+          max_attempts: 5
+          timeout_seconds: 200
+          shell: bash
+          command: |
+            git config --global user.email "csukuangfj@gmail.com"
+            git config --global user.name "Fangjun Kuang"
+
             d=sherpa-onnx-nemo-transducer-giga-am-russian-2024-10-24/
             export GIT_LFS_SKIP_SMUDGE=1
             export GIT_CLONE_PROTECTION_ACTIVE=false
+            rm -rf huggingface
             git clone https://csukuangfj:$HF_TOKEN@huggingface.co/csukuangfj/$d huggingface
-            mv -v $d/* ./huggingface
+            cp -av $d/* ./huggingface
+            cd huggingface
+            git lfs track "*.onnx"
+            git lfs track "*.wav"
+            git status
+            git add .
+            git status
+            git commit -m "add models"
+            git push https://csukuangfj:$HF_TOKEN@huggingface.co/csukuangfj/$d main
+
+      - name: Publish v2 to huggingface (CTC)
+        env:
+          HF_TOKEN: ${{ secrets.HF_TOKEN }}
+        uses: nick-fields/retry@v3
+        with:
+          max_attempts: 5
+          timeout_seconds: 200
+          shell: bash
+          command: |
+            git config --global user.email "csukuangfj@gmail.com"
+            git config --global user.name "Fangjun Kuang"
+
+            d=sherpa-onnx-nemo-ctc-giga-am-v2-russian-2025-04-19/
+            export GIT_LFS_SKIP_SMUDGE=1
+            export GIT_CLONE_PROTECTION_ACTIVE=false
+            rm -rf huggingface
+            git clone https://csukuangfj:$HF_TOKEN@huggingface.co/csukuangfj/$d huggingface
+            cp -av $d/* ./huggingface
             cd huggingface
             git lfs track "*.onnx"
             git lfs track "*.wav"
@@ -145,7 +240,7 @@ jobs:
           HF_TOKEN: ${{ secrets.HF_TOKEN }}
         uses: nick-fields/retry@v3
         with:
-          max_attempts: 20
+          max_attempts: 5
           timeout_seconds: 200
           shell: bash
           command: |
@@ -155,8 +250,9 @@ jobs:
             d=sherpa-onnx-nemo-transducer-giga-am-v2-russian-2025-04-19/
             export GIT_LFS_SKIP_SMUDGE=1
             export GIT_CLONE_PROTECTION_ACTIVE=false
+            rm -rf huggingface
             git clone https://csukuangfj:$HF_TOKEN@huggingface.co/csukuangfj/$d huggingface
-            mv -v $d/* ./huggingface
+            cp -av $d/* ./huggingface
             cd huggingface
             git lfs track "*.onnx"
             git lfs track "*.wav"

scripts/nemo/GigaAM/README.md

@@ -7,4 +7,4 @@ to sherpa-onnx.
 The ASR models are for Russian speech recognition in this folder.
 
 Please see the license of the models at
-https://github.com/salute-developers/GigaAM/blob/main/GigaAM%20License_NC.pdf
+https://github.com/salute-developers/GigaAM/blob/main/LICENSE

scripts/nemo/GigaAM/export-onnx-ctc-v2.py

@@ -1,3 +1,4 @@
+#!/usr/bin/env python3
 import gigaam
 import onnx
 import torch
@@ -27,7 +28,13 @@ def add_meta_data(filename: str, meta_data: dict[str, str]):
 
 def main() -> None:
     model_name = "v2_ctc"
-    model = gigaam.load_model(model_name, fp16_encoder=False, use_flash=False, download_root=".")
+    model = gigaam.load_model(
+        model_name, fp16_encoder=False, use_flash=False, download_root="."
+    )
+
+    # use characters
+    # space is 0
+    # <blk> is the last token
     with open("./tokens.txt", "w", encoding="utf-8") as f:
         for i, s in enumerate(model.cfg["labels"]):
             f.write(f"{s} {i}\n")
@@ -53,5 +60,5 @@ def main() -> None:
     )
 
 
-if __name__ == '__main__':
+if __name__ == "__main__":
     main()

scripts/nemo/GigaAM/export-onnx-ctc.py

@@ -82,6 +82,9 @@ def main():
     model.load_state_dict(ckpt, strict=False)
     model.eval()
 
+    # use characters
+    # space is 0
+    # <blk> is the last token
     with open("tokens.txt", "w", encoding="utf-8") as f:
         for i, t in enumerate(model.cfg.labels):
             f.write(f"{t} {i}\n")

scripts/nemo/GigaAM/export-onnx-rnnt-v2.py

@@ -0,0 +1,158 @@
+#!/usr/bin/env python3
+# Copyright      2025  Xiaomi Corp.        (authors: Fangjun Kuang)
+import os
+
+import gigaam
+import onnx
+import torch
+from gigaam.utils import onnx_converter
+from onnxruntime.quantization import QuantType, quantize_dynamic
+from torch import Tensor
+
+"""
+==========Input==========
+NodeArg(name='audio_signal', type='tensor(float)', shape=['batch_size', 64, 'seq_len'])
+NodeArg(name='length', type='tensor(int64)', shape=['batch_size'])
+==========Output==========
+NodeArg(name='encoded', type='tensor(float)', shape=['batch_size', 768, 'Transposeencoded_dim_2'])
+NodeArg(name='encoded_len', type='tensor(int32)', shape=['batch_size'])
+
+==========Input==========
+NodeArg(name='x', type='tensor(int32)', shape=[1, 1])
+NodeArg(name='unused_x_len.1', type='tensor(int32)', shape=[1])
+NodeArg(name='h.1', type='tensor(float)', shape=[1, 1, 320])
+NodeArg(name='c.1', type='tensor(float)', shape=[1, 1, 320])
+==========Output==========
+NodeArg(name='dec', type='tensor(float)', shape=[1, 320, 1])
+NodeArg(name='unused_x_len', type='tensor(int32)', shape=[1])
+NodeArg(name='h', type='tensor(float)', shape=[1, 1, 320])
+NodeArg(name='c', type='tensor(float)', shape=[1, 1, 320])
+
+==========Input==========
+NodeArg(name='enc', type='tensor(float)', shape=[1, 768, 1])
+NodeArg(name='dec', type='tensor(float)', shape=[1, 320, 1])
+==========Output==========
+NodeArg(name='joint', type='tensor(float)', shape=[1, 1, 1, 34])
+"""
+
+
+def add_meta_data(filename: str, meta_data: dict[str, str]):
+    """Add meta data to an ONNX model. It is changed in-place.
+
+    Args:
+      filename:
+        Filename of the ONNX model to be changed.
+      meta_data:
+        Key-value pairs.
+    """
+    model = onnx.load(filename)
+    while len(model.metadata_props):
+        model.metadata_props.pop()
+
+    for key, value in meta_data.items():
+        meta = model.metadata_props.add()
+        meta.key = key
+        meta.value = str(value)
+
+    onnx.save(model, filename)
+
+
+class EncoderWrapper(torch.nn.Module):
+    def __init__(self, m):
+        super().__init__()
+        self.m = m
+
+    def forward(self, audio_signal: Tensor, length: Tensor):
+        # https://github.com/salute-developers/GigaAM/blob/main/gigaam/encoder.py#L499
+        out, out_len = self.m.encoder(audio_signal, length)
+
+        return out, out_len.to(torch.int64)
+
+    def to_onnx(self, dir_path: str = "."):
+        onnx_converter(
+            model_name=f"{self.m.cfg.model_name}_encoder",
+            out_dir=dir_path,
+            module=self.m.encoder,
+            dynamic_axes=self.m.encoder.dynamic_axes(),
+        )
+
+
+class DecoderWrapper(torch.nn.Module):
+    def __init__(self, m):
+        super().__init__()
+        self.m = m
+
+    def forward(self, x: Tensor, unused_x_len: Tensor, h: Tensor, c: Tensor):
+        # https://github.com/salute-developers/GigaAM/blob/main/gigaam/decoder.py#L110C17-L110C54
+        emb = self.m.head.decoder.embed(x)
+        g, (h, c) = self.m.head.decoder.lstm(emb.transpose(0, 1), (h, c))
+        return g.permute(1, 2, 0), unused_x_len + 1, h, c
+
+    def to_onnx(self, dir_path: str = "."):
+        label, hidden_h, hidden_c = self.m.head.decoder.input_example()
+        label = label.to(torch.int32)
+        label_len = torch.zeros(1, dtype=torch.int32)
+
+        onnx_converter(
+            model_name=f"{self.m.cfg.model_name}_decoder",
+            out_dir=dir_path,
+            module=self,
+            dynamic_axes=self.m.encoder.dynamic_axes(),
+            inputs=(label, label_len, hidden_h, hidden_c),
+            input_names=["x", "unused_x_len.1", "h.1", "c.1"],
+            output_names=["dec", "unused_x_len", "h", "c"],
+        )
+
+
+def main() -> None:
+    model_name = "v2_rnnt"
+    model = gigaam.load_model(
+        model_name, fp16_encoder=False, use_flash=False, download_root="."
+    )
+
+    # use characters
+    # space is 0
+    # <blk> is the last token
+    with open("./tokens.txt", "w", encoding="utf-8") as f:
+        for i, s in enumerate(model.cfg["labels"]):
+            f.write(f"{s} {i}\n")
+        f.write(f"<blk> {i+1}\n")
+        print("Saved to tokens.txt")
+
+    EncoderWrapper(model).to_onnx(".")
+    DecoderWrapper(model).to_onnx(".")
+
+    onnx_converter(
+        model_name=f"{model.cfg.model_name}_joint",
+        out_dir=".",
+        module=model.head.joint,
+    )
+    meta_data = {
+        # vocab_size does not include the blank
+        # we will increase vocab_size by 1 in the c++ code
+        "vocab_size": model.cfg["head"]["decoder"]["num_classes"] - 1,
+        "pred_rnn_layers": model.cfg["head"]["decoder"]["pred_rnn_layers"],
+        "pred_hidden": model.cfg["head"]["decoder"]["pred_hidden"],
+        "normalize_type": "",
+        "subsampling_factor": 4,
+        "model_type": "EncDecRNNTBPEModel",
+        "version": "2",
+        "model_author": "https://github.com/salute-developers/GigaAM",
+        "license": "https://github.com/salute-developers/GigaAM/blob/main/LICENSE",
+        "language": "Russian",
+        "is_giga_am": 1,
+    }
+
+    add_meta_data(f"./{model_name}_encoder.onnx", meta_data)
+    quantize_dynamic(
+        model_input=f"./{model_name}_encoder.onnx",
+        model_output="./encoder.int8.onnx",
+        weight_type=QuantType.QUInt8,
+    )
+    os.rename(f"./{model_name}_decoder.onnx", "decoder.onnx")
+    os.rename(f"./{model_name}_joint.onnx", "joiner.onnx")
+    os.remove(f"./{model_name}_encoder.onnx")
+
+
+if __name__ == "__main__":
+    main()

scripts/nemo/GigaAM/export-onnx-rnnt.py

@@ -83,6 +83,7 @@ def main():
     model.load_state_dict(ckpt, strict=False)
     model.eval()
 
+    # use bpe
     with open("./tokens.txt", "w", encoding="utf-8") as f:
         for i, s in enumerate(model.joint.vocabulary):
             f.write(f"{s} {i}\n")
@@ -94,7 +95,9 @@ def main():
     model.joint.export("joiner.onnx")
 
     meta_data = {
-        "vocab_size": model.decoder.vocab_size,  # not including the blank
+        # not including the blank
+        # we increase vocab_size in the C++ code
+        "vocab_size": model.decoder.vocab_size,
         "pred_rnn_layers": model.decoder.pred_rnn_layers,
         "pred_hidden": model.decoder.pred_hidden,
         "normalize_type": "",

scripts/nemo/GigaAM/run-ctc-v2.sh

@@ -5,11 +5,14 @@ set -ex
 function install_gigaam() {
   curl https://bootstrap.pypa.io/get-pip.py -o get-pip.py
   python3 get-pip.py
+  pip install torch==2.4.0 torchaudio==2.4.0 -f https://download.pytorch.org/whl/torch_stable.html
+  pip install -qq wget text-unidecode "matplotlib>=3.3.2" onnx onnxruntime==1.17.1 pybind11 Cython einops kaldi-native-fbank soundfile librosa
 
   BRANCH='main'
   python3 -m pip install git+https://github.com/salute-developers/GigaAM.git@$BRANCH#egg=gigaam
 
   python3 -m pip install -qq kaldi-native-fbank
+  pip install numpy==1.26.4
 }
 
 function download_files() {

scripts/nemo/GigaAM/run-ctc.sh

@@ -9,7 +9,7 @@ function install_nemo() {
 
   pip install torch==2.4.0 torchaudio==2.4.0 -f https://download.pytorch.org/whl/torch_stable.html
 
-  pip install -qq wget text-unidecode matplotlib>=3.3.2 onnx onnxruntime pybind11 Cython einops kaldi-native-fbank soundfile librosa
+  pip install -qq wget text-unidecode "matplotlib>=3.3.2" onnx onnxruntime==1.17.1 pybind11 Cython einops kaldi-native-fbank soundfile librosa
   pip install -qq ipython
 
   # sudo apt-get install -q -y sox libsndfile1 ffmpeg python3-pip ipython

scripts/nemo/GigaAM/run-rnnt-v2.sh

@@ -0,0 +1,29 @@
+#!/usr/bin/env bash
+# Copyright    2025  Xiaomi Corp.        (authors: Fangjun Kuang)
+
+set -ex
+
+function install_gigaam() {
+  curl https://bootstrap.pypa.io/get-pip.py -o get-pip.py
+  python3 get-pip.py
+  pip install torch==2.4.0 torchaudio==2.4.0 -f https://download.pytorch.org/whl/torch_stable.html
+  pip install -qq wget text-unidecode "matplotlib>=3.3.2" onnx onnxruntime==1.17.1 pybind11 Cython einops kaldi-native-fbank soundfile librosa
+
+  BRANCH='main'
+  python3 -m pip install git+https://github.com/salute-developers/GigaAM.git@$BRANCH#egg=gigaam
+
+  python3 -m pip install -qq kaldi-native-fbank
+  pip install numpy==1.26.4
+}
+
+function download_files() {
+  curl -SL -O https://huggingface.co/csukuangfj/tmp-files/resolve/main/GigaAM/example.wav
+  curl -SL -O https://github.com/salute-developers/GigaAM/blob/main/LICENSE
+}
+
+install_gigaam
+download_files
+
+python3 ./export-onnx-rnnt-v2.py
+ls -lh
+python3 ./test-onnx-rnnt.py

scripts/nemo/GigaAM/run-rnnt.sh

@@ -9,7 +9,7 @@ function install_nemo() {
 
   pip install torch==2.4.0 torchaudio==2.4.0 -f https://download.pytorch.org/whl/torch_stable.html
 
-  pip install -qq wget text-unidecode matplotlib>=3.3.2 onnx onnxruntime pybind11 Cython einops kaldi-native-fbank soundfile librosa
+  pip install -qq wget text-unidecode "matplotlib>=3.3.2" onnx onnxruntime==1.17.1 pybind11 Cython einops kaldi-native-fbank soundfile librosa
   pip install -qq ipython
 
   # sudo apt-get install -q -y sox libsndfile1 ffmpeg python3-pip ipython

scripts/nemo/GigaAM/test-onnx-ctc.py

@@ -19,7 +19,7 @@ def create_fbank():
     opts.frame_opts.window_type = "hann"
 
     # Even though GigaAM uses 400 for fft, here we use 512
-    # since kaldi-native-fbank only support fft for power of 2.
+    # since kaldi-native-fbank only supports fft for power of 2.
     opts.frame_opts.round_to_power_of_two = True
 
     opts.mel_opts.low_freq = 0

scripts/nemo/GigaAM/test-onnx-rnnt.py

@@ -20,7 +20,7 @@ def create_fbank():
     opts.frame_opts.window_type = "hann"
 
     # Even though GigaAM uses 400 for fft, here we use 512
-    # since kaldi-native-fbank only support fft for power of 2.
+    # since kaldi-native-fbank only supports fft for power of 2.
     opts.frame_opts.round_to_power_of_two = True
 
     opts.mel_opts.low_freq = 0
@@ -166,12 +166,7 @@ class OnnxModel:
         target = torch.tensor([[token]], dtype=torch.int32).numpy()
         target_len = torch.tensor([1], dtype=torch.int32).numpy()
 
-        (
+        (decoder_out, decoder_out_length, state0_next, state1_next,) = self.decoder.run(
-            decoder_out,
-            decoder_out_length,
-            state0_next,
-            state1_next,
-        ) = self.decoder.run(
             [
                 self.decoder.get_outputs()[0].name,
                 self.decoder.get_outputs()[1].name,
@@ -213,8 +208,12 @@ def main():
     id2token = dict()
     with open("./tokens.txt", encoding="utf-8") as f:
         for line in f:
-            t, idx = line.split()
+            fields = line.split()
-            id2token[int(idx)] = t
+            if len(fields) == 1:
+                id2token[int(fields[0])] = " "
+            else:
+                t, idx = fields
+                id2token[int(idx)] = t
 
     fbank = create_fbank()
     audio, sample_rate = sf.read("./example.wav", dtype="float32", always_2d=True)